A Critical Role of the C-terminal Segment for Allosteric Inhibitor-induced Aberrant Multimerization of HIV-1 Integrase*

Background: Allosteric integrase (IN) inhibitors (ALLINIs) promote aberrant protein multimerization. Results: ALLINI-2 induces protein-protein interactions, including C-terminal residues Lys-264 and Lys-266, which lead to aberrant, higher-order IN multimerization. Conclusion: The protein-protein contacts beyond the inhibitor binding site contribute to aberrant IN multimerization. Significance: Our findings provide structural clues and underscore the significance for exploiting IN multimerization as a new therapeutic target. Allosteric HIV-1 integrase (IN) inhibitors (ALLINIs) are a promising class of antiretroviral agents for clinical development. Although ALLINIs promote aberrant IN multimerization and inhibit IN interaction with its cellular cofactor LEDGF/p75 with comparable potencies in vitro, their primary mechanism of action in infected cells is through inducing aberrant multimerization of IN. Crystal structures have shown that ALLINIs bind at the IN catalytic core domain dimer interface and bridge two interacting subunits. However, how these interactions promote higher-order protein multimerization is not clear. Here, we used mass spectrometry-based protein footprinting to monitor surface topology changes in full-length WT and the drug-resistant A128T mutant INs in the presence of ALLINI-2. These experiments have identified protein-protein interactions that extend beyond the direct inhibitor binding site and which lead to aberrant multimerization of WT but not A128T IN. Specifically, we demonstrate that C-terminal residues Lys-264 and Lys-266 play an important role in the inhibitor induced aberrant multimerization of the WT protein. Our findings provide structural clues for exploiting IN multimerization as a new, attractive therapeutic target and are expected to facilitate development of improved inhibitors.

HIV-1 integrase (IN) 2 is an important therapeutic target because its function is essential for the virus life cycle. During the early stage of HIV-1 replication, a tetramer of IN assembles with reverse transcribed viral DNA ends to form the stable synaptic complex or intasome. Two catalytic reactions ensue within the nucleoprotein complex. First, a GT dinucleotide is cleaved from the 3Ј terminus of each viral DNA (3Ј-processing). Subsequently, the recessed viral DNA ends are integrated into complementary strands of target DNA in staggered fashion (DNA strand transfer). The integration is facilitated by cellular chromatin-associated protein LEDGF/p75, which tethers the lentiviral intasomes to active genes (1)(2)(3)(4)(5)(6)(7).
Mutagenesis studies have suggested that the role of IN in the virus life cycle extends beyond integration (8 -14). For example, a number of substitutions in the IN coding region have been shown to impair proper maturation of HIV-1 particles and to block the subsequent round of reverse transcription. These mutants have been classified as class II IN mutants as opposed to class I IN mutants, which selectively impair the integration step. Experiments with ectopically expressed Vpr-IN constructs have shown that catalytically defective IN bearing the D116N substitution can complement ␦-IN HIV-1 to rescue reverse transcription in target cells, indicating that the correct IN structure rather than its catalytic activity is important for the late stage of HIV-1 replication (15). However, the exact role of HIV-1 IN during virion maturation remains to be elucidated.
The viral IN protein is composed of three domains (reviewed in Ref. 16). The N-terminal domain (NTD, residues 1-46) contains the zinc-binding HHCC motif, which is essential for the correct folding of this domain and function of the full-length protein (17)(18)(19)(20)(21)(22). The catalytic core domain (CCD, residues 56 -202) contains the DDE active site, which coordinates catalytic Mg 2ϩ ions and mediates catalysis of both 3Ј-processing and strand transfer reactions (23)(24)(25)(26)(27). The C-terminal domain (CTD, residues 220 -288) has a SH3-like fold (28 -30). Each domain is essential for functional protein multimerization and directly interacts with viral DNA. LEDGF/p75 preferentially binds to a tetramer of IN by bridging the CCD-CCD interface of one dimer and the NTD of another dimer (22,31,32). The V-shaped cavity created by two interacting CCDs provides a key docking site for the cellular protein (7). LEDGF/p75 Asp-366 hydrogen bonds with backbone amides of Glu-170 and His-171 of one IN subunit, whereas LEDGF/p75 Ile-365, Leu-368, Phe-406, and Val-408 form hydrophobic contacts with the second IN subunit (7). Additional interactions between LEDGF/ p75 and the IN NTD are mediated by electrostatic interactions (22,32).
Allosteric HIV-1 IN inhibitors (ALLINIs) are an attractive new class of antiviral compounds for clinical development because they target the unexploited IN dimer interface and are expected to complement currently available therapies (reviewed in Ref. [33][34][35][36]. These compounds bind at the IN CCD dimer interface occupying the principal LEDGF/p75 binding pocket (12,(37)(38)(39)(40)(41)(42). Similar to LEDGF/p75 Asp-366, the ALLINI carboxylic acid hydrogen bonds with the main chain nitrogens of residues Glu-170 and His-171 of one IN subunit. The ALLINI methoxy group also hydrogen bonds with the Thr-174 side chain of the same subunit, whereas the ALLINI quinoline ring extends toward another subunit to establish hydrophobic interactions. Consequently, ALLINIs exhibit a dual mode of action: they stabilize interacting IN subunits and promote higher-order protein multimerization as well as inhibit IN-LEDGF/p75 binding in vitro (37,39,43).
In infected cells, ALLINIs impair both the early and late stage of HIV-1 replication underscoring the multifunctional nature of IN during HIV-1 replication (12,39,44,45). When added to target cells, ALLINIs block 3Ј-processing as well as reduce LEDGF/p75-mediated integration into active transcription units (39,46). Knockdown or knock-out of endogenous LEDGF/p75 significantly enhanced the potencies of these inhibitors in target cells, suggesting that there is competition between the cellular protein and ALLINIs for binding to HIV-1 IN during early steps of viral replication (12,47,48). However, the ALLINI antiviral potency is determined primarily through inhibiting the late stage of HIV-1 replication (12,41,42,44,45,48). In producer cells, ALLINIs potently promote aberrant, higher-order multimerization of IN during virus maturation resulting in eccentric, non-infectious cores reminiscent to the phenotype seen with some class II IN mutants (8 -12, 44, 45). In contrast with target cells, the absence of LEDGF/p75 had no effect on ALLINI potencies in the virus producer cells (12,44,47,48). The recent discovery of pyridine-based multimerization selective IN inhibitors, which potently promoted aberrant IN multimerization in vitro and in infected cells but were not effective inhibitors of IN-LEDGF/p75 interactions, have further highlighted IN multimerization as a plausible antiviral target (46). Yet, how a small molecule binding at the IN dimer interface induces aberrant, higher-order protein multimerization is not clear.
HIV-1 genotyping in tissue culture under ALLINI selective pressure has revealed a number of mutations at the inhibitor binding site, including Y99H, L102F, A124D/T124D, A128T, H171T, and T174I substitutions (12, 38 -40, 42, 44). Of these, the A128T substitution has been identified as the most prevalent change and which confers marked resistance to the majority of ALLINIs. Crystallographic studies have shown that a rep-resentative inhibitor, ALLINI-2, interacted with both WT and A128T CCDs (38). The A128T substitution subtly relocated the ALLINI-2 quinoline ring in the A128T CCD dimer co-crystal, whereas the ALLINI-2 carboxylic acid remained similarly hydrogen bonded to WT and A128T CCDs. However, size exclusion chromatography (SEC) experiments with full-length recombinant proteins have revealed that ALLINI-2 induced aberrant multimerization of WT but not A128T IN (38). Consequently, HIV-1 bearing the A128T IN substitution was remarkably resistant to ALLINI-2 (38). How ALLINI-2 binding at the CCD dimer interface promoted differential multimerization of full-length WT and A128T INs remains to be elucidated.
Here, we have used MS-based protein footprinting to monitor ALLINI-2 induced surface topology changes in full-length WT and A128T INs. Our studies show that ALLINI-2 binding promotes protein-protein interactions that extend beyond the direct inhibitor binding site and lead to aberrant multimerization of WT but not A128T IN. These findings further our understanding of HIV-1 IN multimerization as a therapeutic target and will facilitate development of improved antiretroviral agents.

EXPERIMENTAL PROCEDURES
Proteins and the Inhibitor-Full-length recombinant WT and mutant (A128T and K264A/K266A) INs containing the N-terminal His 6 tag were expressed in Escherichia coli and purified as described previously (49). Similar purification protocols were used to prepare isolated IN domains (CCD, CTD, and the CCD-CTD). ALLINI-2 was synthesized as described previously (38).
Differential Hydrogen/Deuterium Exchange (HDX) MS-Solution-phase amide HDX experiments were carried out with a fully automated system (CTC HTS PAL, LEAP Technologies, Carrboro, NC; housed inside a 4°C cabinet) as described previously (50) with slight modifications. 10 M of the His 6 -HIV-1 IN (WT or A128T) protein was mixed with 1:5 molar excess of ALLINI-2 and incubated for 2 h on ice before subjecting to HDX. For the differential HDX experiments, 5 l of either the His 6 -HIV-1 IN (apo) or the complex (1:5 molar mixture of IN and ALLINI-2) were mixed with 20 l of D 2 O-containing HDX buffer (50 mM HEPES 7.5,1 M NaCl, 3 mM DTT) and incubated at 4°C for 0 s, 10 s, 30 s, 60 s, 900 s, or 3,600 s. Following the on exchange reaction, unwanted forward or back exchange was minimized, and the protein was denatured by the addition of 25 l of a quench solution (1% (v/v) TFA in 5 M urea and 50 mM TCEP). Samples were then passed through an immobilized pepsin column (prepared in-house) at 50 l min Ϫ1 (0.1% (v/v) TFA, 15°C), and the resulting peptides were trapped on a C 18 trap column (Hypersil Gold, Thermo Fisher). The bound peptides were then gradient-eluted (5-50% CH 3 CN (w/v) and 0.3% (w/v) formic acid) across a 1-mm ϫ 50-mm C 18 HPLC column (Hypersil Gold, Thermo Fisher) for 5 min at 4°C). The eluted peptides were then subjected to electrospray ionization directly coupled to a high resolution Orbitrap mass spectrometer (LTQ Orbitrap XL with electron transfer dissociation, Thermo Fisher). Each HDX experiment was carried out in triplicate.
Peptide Identification and HDX Data Processing-MS/MS experiments were performed with a LTQ linear ion trap mass spectrometer (LTQ Orbitrap XL with ETD, Thermo Fisher) over a 70-min gradient. Product ion spectra were acquired in a data-dependent mode and the five most abundant ions were selected for the product ion analysis. The MS/MS raw data files were converted to .mgf files and then submitted to Mascot (Matrix Science, London, UK) for peptide identification. Peptides included in the peptide set used for HDX detection had a MASCOT score of 20 or greater. The MS/MS MASCOT search was also performed against a decoy (reverse) sequence, and false positives were ruled out. The MS/MS spectra of all the peptide ions from the MASCOT search were further manually inspected, and only the unique charged ions with the highest MASCOT score were used in estimating the sequence coverage. The intensity weighted average m/z value (centroid) of each peptide isotopic envelope was calculated with the latest version of our in-house developed software, HDX Workbench (51).
MS-based Footprinting of Surface Accessible Lysines-The surface-accessible Lys residues were monitored as described previously (31,52).     (51), and these regions are colored gray. The regions that revealed statistically significant reduction in deuterium uptake in the presence of added inhibitor are colored green and blue according to the heat map coloring scheme used by the software HDX Workbench. The software employs a statistical (p value Ͻ 0.05 for two consecutive time points or a p value Ͻ 0.01 for a single time point) based coloring scheme to distinguish real protection differences from inherent variation in the data. Additionally, overlapping peptides covering the same region and showing a similar protection trend are used to rule out data ambiguity. NS, not significant.
rescence signal was recorded after the addition of europiumstreptavidin using a Perkin Elmer EnSpire multimode plate reader.

RESULTS
Two complementary MS-based approaches were used to identify the inhibitor induced surface topology changes in WT and A128T INs. Amide HDX-MS allowed us to monitor alterations in hydrogen bond networks and conformational changes in the protein backbone, whereas reactivity of side chain primary amines were probed with small molecule modifier NHSbiotin. ALLINI-2 was chosen for these experiments as a representative ALLINI because the crystal structures of its binding to WT and A128T IN CCD dimers are available (38) and allowed us to compare the atomic structures with protein footprinting results for full-length HIV-1 WT and A128T IN interactions with the inhibitor.
Differential rates at which an amide hydrogen exchanges with solvent deuterium can reveal information about inhibitor induced protein conformational changes. Protection from D 2 O uptake results in slower exchange and indicates stronger amide hydrogen bonding suggesting that the protein region has been stabilized and is less dynamic, plastic, or flexible. Conversely, deprotection results in increased exchange, indicating that some of hydrogen bonds are disrupted, suggesting that the protein region is more flexible. Representative HDX results for ALLINI-2 interactions with WT and A128T INs are depicted in Fig. 1, and the data are summarized in Fig. 2. Near complete coverage of the HIV-1 IN primary sequence was obtained enabling detailed analysis of the full-length protein (Fig. 2). For example, significantly slower HDX rates were observed for a representative peptide fragment encompassing regions 155-178 when ALLINI-2 was added to WT or A128T INs, indicating increased protection of this protein segment against D 2 O uptake in the presence of the inhibitor as compared with apo-IN (Fig. 1, A and B, also indicated by the lower black side bar in Fig. 2). This region encompasses the binding pocket for ALLINIs and is in agreement with our crystal structures that show that ALLINI-2 carboxylic acid similarly hydrogen bonds with Glu-170 and His-171 backbone amides of subunit 1 in both the WT and A128T IN CCDs (38). In addition, the substituted benzene ring of the inhibitor engages in hydrophobic interactions with Tyr-99 of subunit 1 in the crystal structure (38). Accordingly, HDX experiments reveal increased protection for the peptides spanning residues 86 -101 (Fig. 2, indicated by the top black side bar) suggesting direct contacts with the inhibitor. The crystal structure of WT CCD bound to ALLINI-2 also showed that the quinoline ring extends toward Ala-128 to establish additional hydrophobic interactions with subunit 2. However, the substitution of Ala with the relatively bulkier and polar Thr relocates the inhibitor away from this position. In keeping with these structural observations, ALLINI-2 binding induces increased protection in the region 121-131 of the WT IN but not the A128T IN when compared with their respective apo-INs (Fig. 2, indicated by the middle black side bar). Taken together, the inhibitor binding patterns observed with the WT and A128T IN CCDs crystal structures were closely recapitulated using HDX in the context of fulllength proteins.
Importantly, HDX experiments have identified additional IN segments affected by the inhibitor binding to full-length IN that extend beyond the direct ALLINI-2 binding sites seen in crystal structures with the CCD. The peptides spanning amino acids 224 -248 were similarly affected by the inhibitor binding to WT and A128T INs (Fig. 2, indicated by the middle magenta side bar). However, two IN segments: 181-202 and 250 -280, were uniquely affected by the inhibitor binding to WT IN but not the A128T variant (Fig. 2, indicated by top and bottom magenta side bars). For instance, a peptide overlapping residues 257-288 in the WT IN in complex with ALLINI-2 displays increased protection in comparison to the same peptide from its unliganded counterpart (Fig. 1C). However, HDX rates for the identical peptide remained very sim-ilar when A128T IN was compared in the presence and absence of ALLINI-2 (compare Fig. 1, C and D).
We next extended footprinting of WT and A128T INs to compare reactivity of Lys residues to a small molecule modifier, NHS-biotin, in the presence and absence of ALLINI-2 (31,52). Of particular interest are the residues that are susceptible to modifications in free protein but are shielded from modification upon complex formation. Treatment of apo-INs with NHS-biotin revealed a very similar modification patterns for WT and A128T INs, indicating that the single amino acid substitution did not detectably alter the overall structure of IN. Ten modified Lys residues were detected in each protein, including Lys-14, Lys-159, Lys-160, Lys-186, Lys-215, Lys-219, Lys-244, Lys-258, Lys-264, and Lys-266. Of these, only a single tryptic peptide peak (264 -269) containing biotinylated Lys-264 and Lys-266 was markedly diminished upon ALLINI-2 addition to WT IN (Fig. 3, compare B and C). In contrast, only a slight reduction of this peak was observed when ALLINI-2 was added to A128T IN (Fig. 3, compare E and F). These findings were highly reproducible in three independent experiments. Reactivity levels of the other nine Lys residues to NHS-biotin were not significantly affected when WT and A128T INs were compared in the absence or presence of ALLINI-2. As a representative data, the modification patterns for the peptide 216 -224 containing modified Lys-219, which remained reactive to NHS-biotin in the presence and absence of the inhibitor is shown in Fig. 3. Collectively, the findings illustrated in Fig. 3 agree with and extend the HDX data (Fig. 1B), implicating the 257-280 segment and in particular Lys-264 and Lys-266 in the inhibitor induced structural changes seen in WT but not A128T IN structures.
The structural changes observed in the two domains (the CCD and CTD) could be due to either direct ALLINI-2 binding to IN or inhibitor induced protein-protein interactions. To delineate these possibilities, we have examined ALLINI-2 bind-ing to the isolated WT and A128T CCDs as well as to WT CTD and the two domain CCD-CTD using SPR (Fig. 4). The inhibitor bound to WT and A128T CCDs with comparable K d values of 0.722 Ϯ 0.2 M and 1.66 Ϯ 0.2 M, respectively (Fig. 4, A and  B), indicating that the A128T substitution did not significantly affect ALLINI-2 binding. The inhibitor bound to WT CCD-CTD with a K d of 2.64 Ϯ 0.4 M (Fig. 4C), whereas it failed to bind the WT CTD (Fig. 4D), suggesting that the CTD does not contribute to ALLINI-2 binding. Although poor solubility of full-length WT IN does not permit its analysis by SPR, it should be noted that the K d values obtained with the isolated WT CCD and CCD-CTD domains correlate well with the ALLINI-2 IC 50 of ϳ0.6 M against WT HIV-1 (38). Taken together, the data in Fig. 4 along with available structural data (38) help to interpret our MS-based protein footprinting results. It is logical to conclude that the IN CCD segments indicated by black side bars in Fig. 2 are affected by direct binding of ALLINI-2 to IN, whereas the additional regions indicated by magenta side bars in Fig. 2 reflect protein-protein interactions induced by the inhibitor binding.
We next examined ALLINI-2 induced HIV-1 IN multimerization using DLS (Fig. 5). This technique can detect particle aggregation and has enabled us to identify higher-order oligomerization of IN in the presence of multimerization selective pyridine-based inhibitor KF116 (46). In the absence of ALLINI-2, no aggregation of apo-IN (200 nM) was observed during the entire time course of the experiment. The lower oligomeric soluble forms (monomers, dimers, and tetramers) of IN were not detected by this technique. ALLINI-2 (200 nM) alone was also soluble in the same buffer. In contrast, light scattering recorded 1, 5, and 15 min after addition of ALLINI-2 to WT IN revealed peaks corresponding to particle sizes with diameters of 164, 342, and 615 nm, respectively (Fig. 5A), indicating an equilibrium shift toward higher-order oligomers in a time-dependent manner. Of note, these particle sizes are significantly larger than functional HIV-1 IN tetramers (with a diameter of ϳ7.5 nm) seen by atomic force microscopy analysis of in vitro-assembled HIV-1 intasomes (53). We have also performed DLS experiments with IN containing either A128T (Fig.  5B) or K264A/K266A (Fig. 5C) substitutions. Strikingly, both mutant proteins exhibited remarkable resistance to ALLINI-2induced protein multimerization.
The effects of A128T and K264A/K266A substitutions were also examined by SEC. Here, tetramers and monomers were seen for WT and the mutant proteins. Addition of ALLINI-2 to WT IN markedly reduced the tetrameric form. However, higher-order oligomers seen in the DLS experiments could not be detected by SEC possibly due to the inability of large protein aggregates to migrate through the Superdex 200 matrix. In contrast, ALLINI-2 addition to A128T or K264A/K266A IN did not significantly alter the tetrameric or monomeric species.
To further examine the effects of A128T and K264A/K266A substitutions on IN structure and function, we carried out additional biochemical and biophysical characterization of the mutant proteins. MS-based protein footprinting of WT and the two mutant proteins displayed very similar surface accessibility profiles, indicating that A128T or K264A/K266A substitutions did not significantly alter the protein structure. The K264A/ K266A IN mutant exhibited WT levels of LEDGF/p75 binding, whereas A128T IN displayed slightly reduced binding to the cellular cofactor (Fig. 6B). Catalytic activities of WT and A128T INs were comparable, whereas K264A/K266A IN was not active (Fig. 6C).

DISCUSSION
The crystal structure of ALLINI-2 binding to the isolated CCD dimer has revealed that the inhibitor bridges between two IN subunits (38). However, how these interactions promote aberrant, higher-order IN multimerization is not clear. MS-based protein footprinting studies of full length IN, the results of which are summarized in Fig. 7A in the context of the IN structural model (54), allowed us to identify protein-protein interactions in the CCD and CTD that extend beyond the inhibitor binding sites and contribute to higher-order IN multimerization. A recent study (41), which investigated the multimerization of IN domains in the presence of ALLINI GSK1264 using sedimentation velocity and turbidity assays, has also revealed the significance of the CTD in addition to the CCD for the formation of insoluble protein aggregates. In par-ticular, the addition of GSK1264 stimulated dimerization of IN F185H (CCD) and IN F185H (NTD-CCD), whereas IN F185H (CCD-CTD) as well as full-length IN showed evidence of the inhibitorinduced protein aggregation (41). Our DLS analysis of ALLINI-2 interactions with the WT CCD-CTD truncated mutant did not result in detectable particle aggregates as was seen with full-length WT IN (data not shown). These discrepancies could be due to different experimental conditions employed in the two studies. We show that ALLINI-2 binds WT and the drug-resistant A128T CCDs with comparable affinities (Fig. 4). Yet, WT but not the mutant IN was susceptible to inhibitor-induced aber- rant multimerization (Fig. 5). Available crystal structures provide potential explanation for these observations (38). The hydrogen bonding network between ALLINI-2 and CCD subunit 1, which is the primary determinant of the inhibitor binding, is preserved for both WT and A128T CCDs (38). However, the A128T IN substitution relocates the quinoline ring away from CCD subunit 2, which could reduce the ability of ALLINI-2 to effectively bridge the two subunits and promote higher-order multimerization of the mutant IN. Yet, it is possible that ALLINI-2 can still induce limited conformational changes in A128T IN resulting in protections in the 226 -241 region that were detected with HDX experiments.
We have identified two IN regions 181-202 and 250 -280 uniquely protected in WT but not in the A128T protein, suggesting their role in the inhibitor-induced aberrant IN multimerization. In particular, our findings demonstrate the significance of residues Lys-264 and Lys-266 for higher-order protein multimerization. In the NMR structure of the isolated CTD, which is seen as a dimer with Lys-264 and Lys-266 being fully surface-exposed and positioned on the opposite side of the CTD-CTD interface (28 -30). In the HIV-1 intasome models, which have been generated based on the analogous prototype foamy virus intasome structure, the CTD domains do not engage in protein-protein interactions and instead interact with viral DNA (10,54,55). In particular, Lys-264 and Lys-266 directly contact viral DNA (Fig. 7B). In keeping with these observations, K264A/K266A IN substitutions compromised catalytic activities (Fig. 6C). Furthermore, substitutions introduced at IN residues Lys-264 and Lys-266 markedly impaired HIV-1 replication (56).
In vitro ALLINIs compromise IN catalytic activities by blocking formation of the stable synaptic complex between IN and viral DNA (37). Consistently, ALLINIs impaired a step at or prior to 3Ј-processing in infected cells (39). These observations could be explained by our findings that ALLINI-2-induced IN multimerization compromises the surface exposure of Lys-264 and Lys-266, which are needed for IN binding to viral DNA.
At the same time, it should be noted that ALLINIs are more potent during virus particle maturation rather than during the early steps of HIV-1 replication (12,41,42,44,45). The competition between chromatin associated LEDGF/p75 and ALLINI binding to the HIV-1 IN dimer interface appears to be one significant factor reducing the inhibitor potency in target cells. Such a competition is absent or minimal during the late stage of HIV-1 replication, which allows ALLINIs to potently induce aberrant IN multimerization, leading to the formation of eccentric cores with the electron dense material being mislocalized between the translucent capsid core and particle membrane. Therefore, it would be interesting to examine contributions of the CTD fragments identified in the present study and in particular Lys-264 and Lys-266 to promote aberrant IN multimerization during the virus particle maturation during ALLINI treatment. These studies, which are currently underway in our laboratory, will help to better understand the mechanism of action of available ALLINIs and facilitate development of improved inhibitors.